New Cisco UCS Server Takes Aim At AI With Latest Nvidia GPUs

Matt Broersma is a long standing tech freelance, who has worked for Ziff-Davis, ZDnet and other leading publications

Follow on: Google +

Nvidia’s graphics processing chips can drastically reduce the time needed to train deep learning models, Cisco says

Cisco has expanded its Unified Computer System (UCS) range to include systems optimised for artificial intelligence tasks using Nividia’s latest graphics chips.

The networking giant is the latest company to target offerings at the expanding AI field, which is booming in part due to demand for the technology in analysing vast amounts of data.

Competitors including Dell, HPE and Lenovo are also offering AI-focused systems for sectors including finance, healthcare, media, security, retail and manufacturing.

Cisco said it has seen its big data analytics business grow by a factor of 18 in the past four or five years, with demand ranging from large enterprises to the public sector.

Cisco’s new UCS C480 ML M5. Cisco

GPU power

Its new UCS C480 ML M5 is a 4U-form factor server with Intel Xeon processors and eight Nvidia Tesla V100-32G GPUs with NVLink interconnects.

Users can flexibly configure CPU, networking, storage, memory and software options, with the top-end configuration including dual Xeon chips, up to 128GB of 2666MHz DDR4 RAM, 24 SATA hard drives or SSDs, six NVMe drives and four x100G Virtual Interface Cards.

Cisco said GPUs are particularly helpful in reducing the amount of time it takes to train AI models, in deep learning for instance.

The new rack server is designed to work with Cisco’s other servers and HyperFlex hyperconverged systems.

Cisco has also partnered with third party big data companies for UCS integration, including Cloudera, Hortonworks and MapR. The system also supports frameworks such as TensorFlow and PyTorch.

Cisco said it is working with Hortonworks to validate Hadoop 3.1 in a design where the new UCS server stores data and runs containerised Apache Spark and Google TensorFlow analytic workloads.

The UCS C480 ML M5 Rack Server is set to be available in the fourth quarter.